aserio changed the topic of #ste||ar to: STE||AR: Systems Technology, Emergent Parallelism, and Algorithm Research | stellar.cct.lsu.edu | HPX: A cure for performance impaired parallel applications | github.com/STEllAR-GROUP/hpx | Buildbot: http://rostam.cct.lsu.edu/ | Log: http://irclog.cct.lsu.edu/
parsa has joined #ste||ar
parsa has quit [Read error: Connection reset by peer]
parsa| has joined #ste||ar
<github> [hpx] hkaiser force-pushed find_symbols from 5219aff to c38c6dc: https://git.io/vbzvG
<github> hpx/find_symbols c38c6dc Hartmut Kaiser: Change existing symbol_namespace::iterate to return all data instead of invoking a callback
<K-ballo> is it possible to have ctest output junit xml results, or some other well known format?
diehlpk has joined #ste||ar
<zao> I believe it's rather insistent on emitting its own, at least that's what I found when looking into it.
K-ballo has quit [Quit: K-ballo]
jbjnr has quit [Read error: Connection reset by peer]
hkaiser has quit [Quit: bye]
mcopik has quit [Ping timeout: 268 seconds]
diehlpk has quit [Ping timeout: 272 seconds]
nanashi55 has quit [Ping timeout: 258 seconds]
nanashi55 has joined #ste||ar
parsa| has quit [Quit: Zzzzzzzzzzzz]
kvuser2 has joined #ste||ar
<kvuser2> Hi my name is Adarsh ,I am new to open-source and want to start contributing. What path should I follow to get started. I have programming experience in C++,C and Python. What repositories and what projects would you suggest I should take up
kvuser2 has quit [Quit: Page closed]
jbjnr has joined #ste||ar
<jbjnr> heller: zao et al, warning. Moving pycicle into a separate repo has messed up the paths in some scripts, so I will fix them shortly.
<jbjnr> docs are wrong
<heller> OK
<heller> Thanks
<heller> jbjnr: can the job submission be splitted such that we can have a different number of tasks/nodes for running the tests?
<heller> One node for compiling, two for running the tests, for example
<jbjnr> heller: doable, but might be a bit more complicated than it ought to be to make sure that if someone force pushed a branch, then a new slurm task didn't clobber the working dir. between the original build and the secondary test.
<jbjnr> I didn't do this yet because v1.
<jbjnr> but multinode testing is in my plan
<jbjnr> because libfabric/mpi etc
<heller> Yeah
<heller> Tcp needs proper distributed testing as well
<heller> The setup should be doable with job dependencies
<heller> And we know the name of the jobs...
<jbjnr> yup. we can slowly add features like that once we have FAU/LSU/zao? up and running ...
david_pfander has joined #ste||ar
jaafar_ has quit [Ping timeout: 240 seconds]
<jbjnr> heller: some fixes and doc updates https://github.com/biddisco/pycicle
<jbjnr> I think it should work if following the install steps given
<jbjnr> heller: zao shall I give you write access to the repo so that if either of you try running it you can push tweaks directly to the repo?
<jbjnr> shit. I messed up again. Force pushing
<jbjnr> ok. think it's right now
<github> [hpx] msimberg pushed 1 new commit to gh-pages: https://git.io/vbgKN
<github> hpx/gh-pages 75f70b3 Mikael Simberg: Remove outdated index.html...
<github> [hpx] msimberg deleted maste at 75f70b3: https://git.io/vbgi9
<simbergm> bah, heller do you remember how you generated the index.html in the first place?
<github> [hpx] msimberg pushed 1 new commit to gh-pages: https://git.io/vbgPb
<github> hpx/gh-pages 126daae Mikael Simberg: Set theme jekyll-theme-minimal and migrate Page Generator content
<github> [hpx] msimberg pushed 1 new commit to gh-pages: https://git.io/vbgX4
<github> hpx/gh-pages 7c59cca Mikael Simberg: Remove all automatically generated files
<github> [hpx] msimberg pushed 1 new commit to gh-pages: https://git.io/vbgXz
<github> hpx/gh-pages fa649c7 Mikael Simberg: Set theme jekyll-theme-minimal
<Guest36397> [hpx] msimberg pushed 1 new commit to gh-pages: https://git.io/vbgXi
<Guest36397> hpx/gh-pages e920836 Mikael Simberg: Try adding README.rst from master branch
<simbergm> well, download buttons are gone but at least the readme is up to date now
<simbergm> good enough for now
<heller> simbergm: no, i think it was automated :(
<simbergm> heller: yeah, I suspect so too after having a go at it :) they changed the config files etc so it works a bit differently now
<simbergm> we can update the index.md if/when we set up circleci to do the docs
<simbergm> heller: are you happy to merge 3051?
<simbergm> also, gtg
<heller> simbergm: yes, looks good
parsa has joined #ste||ar
<heller> jbjnr: the dashboard looks messed up regarding builds on master
<jbjnr> ok. I'll take a look. I "simplified"it yesterday. might have broken somethng
<jbjnr> what seems wrong?
<heller> master shows up on the Pull Requests track
<jbjnr> I see it
<jbjnr> yes. caused by my simplification indeed
K-ballo has joined #ste||ar
<jbjnr> pushed fix. master should be ok now. going for lunch
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
hkaiser has joined #ste||ar
<simbergm> hmm, I broke the docs builder on rostam, it needs to do a pull before pushing
<simbergm> but I find no mention of pushing/adding/gh-pages in the hermione-buildbot repo
<simbergm> is that what's used on rostam?
<simbergm> heller? hkaiser?
daissgr has quit [Ping timeout: 264 seconds]
<heller> simbergm: it's in the CMakeLists.txt
<heller> simbergm: make docs && make git_docs
<simbergm> heller: thanks
daissgr has joined #ste||ar
<simbergm> incidentally, I just tried to build hpx with std=c++11 and it breaks, is it supposed to work or is it time to throw out support for that?
<hkaiser> what breaks?
<simbergm> errors are about auto and missing return types, so could be easy to fix
<hkaiser> it's supposed to work
<simbergm> error: ‘operator()’ function uses ‘auto’ type specifier without trailing return type
<simbergm> for example
<hkaiser> where?
<simbergm> sec
<hkaiser> that should be protected by a pp macro
<simbergm> hpx/traits/future_then_result.hpp:73:50
<hkaiser> ok, that's a bug
<simbergm> i only set HPX_WITH_CXX11:BOOL=TRUE and 14 and 17 to false in cmake
<simbergm> that should be enough?
<hkaiser> HPX_HAVE_CXX14_RETURN_TYPE_DEDUCTION seems to be set anyways
<hkaiser> so either you used an inconsistent cmake cache or the tests are broken
<hkaiser> feature tests, that is
<simbergm> okay, I'll try once with a clean one before I create some work for you
<simbergm> :)
<hkaiser> ok
<K-ballo> bad cache
<simbergm> hkaiser: it was indeed a bad cache
<hkaiser> nod
<hkaiser> K-ballo: is the fixed mpark::variant available as a single file version already?
<simbergm> but then changing the std setting on an existing cache is not supported/recommended/broken
<simbergm> or I just did it wrong...
<hkaiser> simbergm: looks like
<hkaiser> but changing existing cmake caches is never safe anyways
<K-ballo> hkaiser: yeah, it already was when I linked it the other day, I had asked for it
<K-ballo> I believe it is now being generated automatically on each commit
<simbergm> so then it's not supported
<hkaiser> ahh, ok - then I misunderstood you the other day
<hkaiser> thanks
<simbergm> thanks
<hkaiser> simbergm: yah
<K-ballo> hkaiser: nah, it wasn't available when I first told you the other day, so I requested it, and mpark delivered within minutes
<hkaiser> k
<hkaiser> K-ballo: btw, we got the go ahead for FLEcsI
<hkaiser> let's schedule a phone call or somesuch early january
<K-ballo> cool
<github> [hpx] msimberg opened pull request #3074: Fix more build warnings (master...fix-build-warnings) https://git.io/vb2fM
david_pfander has quit [Remote host closed the connection]
david_pfander has joined #ste||ar
david_pfander has quit [Remote host closed the connection]
hkaiser has quit [Quit: bye]
aserio has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
parsa has quit [Client Quit]
parsa has joined #ste||ar
parsa has quit [Client Quit]
<zao> This may be relevant for those of you building on Windows - Defender makes process spawning take a nice comfy 260ms longer.
* K-ballo is glad he doesn't have defender
<zao> I didn't quite think this through... I've got the source code and the build images for my builder on a four-disk ZFS stripe, and no machine with enough free SATA ports to mount it.
david_pfander has joined #ste||ar
akheir has joined #ste||ar
akheir has quit [Remote host closed the connection]
aserio1 has joined #ste||ar
hkaiser has joined #ste||ar
hkaiser has quit [Client Quit]
hkaiser has joined #ste||ar
aserio has quit [Ping timeout: 255 seconds]
aserio1 is now known as aserio
<aserio> wash[m]: will you be joining us today?
eschnett has quit [Quit: eschnett]
<github> [hpx] msimberg pushed 3 new commits to master: https://git.io/vb24l
<github> hpx/master 964b573 Bruno Pitrus: Fixed error in documentation for container version of copy.
<github> hpx/master ddbc8f9 Bruno Pitrus: Container version of move
<github> hpx/master f5bfd65 Mikael Simberg: Merge pull request #3028 from brjsp/move_rangests...
<heller> hkaiser: my main point regarding #3072 is the credit splitting. The other comment is more a general one, unrelated to the actual implementation
<github> [hpx] hkaiser pushed 1 new commit to find_symbols: https://git.io/vb24E
<github> hpx/find_symbols a9e0248 Hartmut Kaiser: Addressing review comments
<aserio> zbyerly: will you be joining us today?
<hkaiser> heller: yah I addressed the splitting issue
<heller> In general, I'm just sceptical to those global operations
<hkaiser> heller: any better ideas?
<heller> No, as I'm not sure what the final application for phylanx is... But I'm almost sure, that the broadcast based approach will not be the most optimal one. As an API for exposing the information, it's fine though
<heller> Only for displaying the names on stdout, it's definitely not as good as the call back based solution
<heller> Especially since you should be able to derive the names from the ast
<hkaiser> heller: not necessarily
<hkaiser> well, that might be possible, but would require duplicating most of the compiler code
<heller> Or saving the names on the node
<heller> in*
<heller> Which the client should do already
<hkaiser> heller: that means replicating a symbolic namespace in the app
<hkaiser> I'm not worried about performance of the compilation in phylanx, currently - so I think I'll be fine
<heller> hkaiser: I don't think so, you still use the symbolic namespace for name resolution
<hkaiser> sure
<heller> But then yeah, you might as well store the gids in the ast
<heller> (that's me not knowing how phylanx works in distributed)
<hkaiser> I want to have globally discoverable symbols anyways
<hkaiser> for viz and analysis purposes
david_pfander has quit [Ping timeout: 272 seconds]
mcopik has joined #ste||ar
aserio has quit [Ping timeout: 255 seconds]
aserio has joined #ste||ar
K-ballo has quit [Quit: K-ballo]
<heller> hkaiser: sure. Just bouncing off ideas without knowing too much about the actual use case
<heller> hkaiser: another approach using the callback based design would be to use a channel +async iterators instead of storing everything
K-ballo has joined #ste||ar
jaafar_ has joined #ste||ar
aserio has quit [Ping timeout: 240 seconds]
mcopik has quit [Ping timeout: 272 seconds]
Smasher has joined #ste||ar
aserio has joined #ste||ar
mcopik has joined #ste||ar
aserio has quit [Ping timeout: 255 seconds]
<nanashi55> hkaiser: My paper about HPX is almost finished and I'm going to submit it on monday. Do you want to read the draft first or do you just want to see the submitted version?
<hkaiser> heller: using channels is an interesting idea
hkaiser has quit [Read error: Connection reset by peer]
parsa has joined #ste||ar
hkaiser has joined #ste||ar
<hkaiser> nanashi55: I'd love have a look at what you have right now
<nanashi55> Great. Can I send it to you by email?
aserio has joined #ste||ar
<hkaiser> nanashi55: sure: hartmut.kaiser@gmail.com
parsa has quit [Quit: Zzzzzzzzzzzz]
<heller> what's the paper about?
<nanashi55> heller: It's just a quick overview of HPX
<heller> I see
<github> [hpx] sithhell pushed 1 new commit to master: https://git.io/vb2S3
<github> hpx/master 7575200 Thomas Heller: Merge pull request #3071 from STEllAR-GROUP/fix_pack_traversal_test...
<nanashi55> hkaiser: sent
<hkaiser> nanashi55: got it, thanks
<heller> hkaiser: sorry, the credit splitting won't work that way
hkaiser has quit [Quit: bye]
hkaiser has joined #ste||ar
<hkaiser> aserio: yt?
<aserio> hkaiser: yes
<hkaiser> just sent an email wrt scheduling a call tomorrow or friday
<hkaiser> could you take care of that, pls?
<aserio> hkaiser: sure
<hkaiser> aserio: tks
<aserio> hkaiser: is there a reason VS doesn't like floats
<hkaiser> no reason, what is it complaining about?
<aserio> narrowing
<aserio> I am assigning a float to a number like 10.01
<aserio> I assume it would prefer me to use doubles
<hkaiser> ahh, initializer lists prevent integral and floating point types from narrowing conversions
<hkaiser> use 10.1f to specify a floating point constant
kazbek has joined #ste||ar
<kazbek> I have troubles installing HPX on OS X
<heller> kazbek: tell me more
<heller> kazbek: i literally saw a person compiling and running a HPX application on his macbook this monday
<kazbek> I used brew to install missing packages: gcc-7, jemalloc, hwloc, mpich
<heller> ok
<kazbek> cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/usr/local -DHPX_WITH_PARCELPORT_MPI=true -DHPX_WITH_MALLOC=jemalloc -DHPX_WITH_THREAD_IDLE_RATES=true -DHPX_WITH_CXX14=on -DHPX_WITH_TESTS=off -DHPX_WITH_EXAMPLES=off -DCMAKE_CXX_COMPILER=/usr/local/Cellar/gcc/7.2.0/bin/g++-7 -DCMAKE_C_COMPILER=/usr/local/Cellar/gcc/7.2.0/bin/gcc-7 ..
<kazbek> that the cmake run
<kazbek> and I get error
<kazbek> Linking CXX shared library ../lib/libhpx.dylib Undefined symbols for architecture x86_64:
<heller> hmm
<heller> I know that it works with appleclang
<kazbek> i tried it too
<kazbek> used native clang
<heller> which cmake version?
<kazbek> it gave the same error
<kazbek> 3.10
<kazbek> my os x version is 10.13
<kazbek> high sierra
<heller> which undefined symbols is it complaining about?
<kazbek> [ 84%] Linking CXX shared library ../lib/libhpx.dylib Undefined symbols for architecture x86_64: "_MPI_Barrier", referenced from: hpx::parcelset::parcelport_impl<hpx::parcelset::policies::mpi::parcelport>::stop(bool) in parcelport_mpi.cpp.o "_MPI_Comm_dup", referenced from: hpx::util::mpi_environment::init(int*, char***, hpx::util::command_line_handling&) in mpi_environment.cpp.o
<kazbek> and many more
<kazbek> but all of them are MPI related
<heller> looks like an MPI problem then
<heller> if you don't have a OSX cluster, it is safe to turn it of for your build on your local machine
<kazbek> I tried to compile a code with mpi
<kazbek> and it worked allright
<heller> *shrug*
<heller> we might miss some libraries that need to be linked in
<kazbek> not use PARCELPORT_MPI?
<heller> on your local machine, yes
<kazbek> ok
<heller> not needed there
<kazbek> allright
<kazbek> I'll give it a shot. Thanks!
<heller> can't help you debug the MPI problem (i don't have access to a OSX machine...)
<heller> that's related, I guess
<kazbek> it built allrigh without the parcelport
<kazbek> so it will be sufficient to run it on my machine with mpirun?
<kazbek> this is just for developing purposes
<heller> Use hpxrun.py
<heller> This will give you a good start
kazbek has quit [Quit: Page closed]
zao has quit [Quit: Up, up, and away!]
zao has joined #ste||ar
Smasher has quit [Remote host closed the connection]
Smasher has joined #ste||ar
daissgr1 has joined #ste||ar
thundergroudon[m has quit [Ping timeout: 240 seconds]
thundergroudon[m has joined #ste||ar
aserio has quit [Quit: aserio]
Smasher has quit [Remote host closed the connection]
EverYoung has joined #ste||ar